Set the w_pecision = 8 in the gen_weights.py and C-Simulation Fail

[lkl110137918218]

Hi.
I am currently trying to use the conv layer in this library (i.e. test_conv3.tcl) and testing it on the first convolution layer of VGG16. When I set the value below in the gen_weights.py.

• kernel_dim = 3
• stride = 1
• input_precision = 8
• ifm_channels = 3
• ofm_channels = 64
• ifm_dimension = 226
• ofm_dimension = 224
• activation_precision = 16
• expand = 1 (I'm not sure what this is for)
• simd = 3
• pe = 4
• w_precision = 8
• mmv=1 (I'm not sure what this is for)
  It failed during the C-simulation, as shown in the figure below, with a difference between the expected and actual values.
  

BUT, If I change the w_precision, i.e., w_precision = 1, everything works well, and C-simulation is successful. I have no idea how to fix this problem because I only set the different parameter at w_precision and didn't modify any source code. Please help me, thanks.